1. Because finding nodal points, etc. is a pain, and because pano rigs tend to be big and cumbersome to set up.
2. Because sometimes the stitching process for a spherical pano just fails and you can't put the image together. I've never had a shifted pano fail to stitch.
3. Because you can't easily use filters with a spherical pano, since the lens moves. When doing a shifted pano, you can use graduated ND filters (useful if there's a moving object in one part of the scene) and polarisers.
4. Because, even if the centre of a prime is sharper than the corners of a tilt-shift lens, the distortion induced by correcting the spherical pano into a rectilinear projection will make the corners of the final image worse than the corners of a shifted image, when you're dealing with the wider focal lengths.
1. Euh not really. Finding the nodal point of a lens is a simple process performed only once. Pano heads are a bit bulky, but the good ones fold into a compact package. By the way the pano above was done without a pano head, not that I would recommend this as a best practise,
2. Never happened to me after thousands of spherical panos, not a single time. When using pano heads I also don't remember many occurences where blending issues occured, just a few with sea scenes at fast shutter speeds (zero issues when using slow shutter speeds obviously),
3. I can relate to what you are saying only for neutral grads if you do multi-rows panos... but I don't see any need for those filters in the first place with modern cameras like the D800. PL is not an issue, at least no more than with other wide angle options (meaning mostly a bad idea). But spherical pano has major advantages with flare control when using filters since the longer lens you are using typically have more protective hoods, this is also a major value when shooting in the rain for example,
4. Not with modern software like PT gui. There is of course an angular coverage beyond which a planar projection becomes impossible, but the information available on the faceted surface of the sphere of a mosaic stitch is not significantly less in the corners vs the center compared to the mosaic after projection on a plane. What is less, of course, is the amount of pixels available to cover the same angular section of the scene in the corners vs the center... but that is exactly the same with a single lens. If you haven't I encourage you to do this test by yourself, a 24mm is a good candidate.
So, I am sorry, but I am still not convinced.